Update audioset.py

audioset.py

@@ -16,6 +16,8 @@ from huggingface_hub import hf_hub_download
 
 SAMPLE_RATE = 22_050
 
+_HOMEPAGE = "https://huggingface.co/datasets/confit/audioset"
+
 _BALANCED_TRAIN_FILENAME = 'balanced_train_segments.zip'
 _EVAL_FILENAME = 'eval_segments.zip'
 
@@ -89,16 +91,16 @@ class AudioSet(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, archive_path, split=None):
         extensions = ['.wav']
         
-        test_metadata_csv = os.path.join('metadata/eval_segments.csv')
+        test_metadata_csv = f"{_HOMEPAGE}/resolve/main/metadata/eval_segments.csv"
         if self.config.name == 'balanced':
-            train_metadata_csv = os.path.join('metadata/balanced_train_segments.csv')
+            train_metadata_csv = f"{_HOMEPAGE}/resolve/main/metadata/balanced_train_segments.csv"
         elif self.config.name == 'unbalanced':
-            train_metadata_csv = os.path.join('metadata/unbalanced_train_segments.csv')
+            train_metadata_csv = f"{_HOMEPAGE}/resolve/main/metadata/unbalanced_train_segments.csv"
         train_metadata_df = self._preprocess_metadata_csv(train_metadata_csv) # ['filename', 'ids']
         test_metadata_df = self._preprocess_metadata_csv(test_metadata_csv) # ['filename', 'ids']
         
         class_labels_indices_df = pd.read_csv(
-            os.path.join('metadata/class_labels_indices.csv')
+            f"{_HOMEPAGE}/resolve/main/metadata/class_labels_indices.csv"
         ) # ['index', 'mid', 'display_name']
         mid2label = {
             row['mid']:row['display_name'] for idx, row in class_labels_indices_df.iterrows()